How to read this desk, and the ground under it

If all three name the same session, the page is coherent and there is nothing to think about. If the index clock is newer than the chain clock, that is the ordinary afternoon state: the exchange publishes the close at 15:30 and the full option file lands in the evening, so between the bell and the files the index has moved on and the chain has not. The block says so rather than hiding it.

It is not a claim that the desk is current. When a session has closed and the desk has no row for it, a caution line appears here saying exactly that. A page with no caution and one date is a page whose figures all belong to the same market.

Read the outcome column as a record and not as a grade. Roughly fifty-five per cent of breaks on this archive closed back through the level within five sessions, which is the useful thing to know about them.

It is not a signal. Whether daily breaks follow through was tested against a shuffle null on both indices: NIFTY came in below random ground at every horizon, one cell inverted, and BANKNIFTY was not significant. Whether the height of the formation before a break says how far it goes was tested too, and the correlation was about zero on both indices at every horizon.

The payoff at expiry is exact arithmetic on the legs. The payoff before expiry is a model, and the two are separately named functions so one is never handed over as the other; where the model cannot be evaluated the curve is absent rather than falling back to the expiry shape under the same label.

It proposes nothing and ranks nothing. The shapes are named for what they are and are not filed under bullish, bearish or neutral, because that filing is an opinion about the market wearing a filing system. Where a position's loss is unbounded it is reported as unbounded rather than as the number at the edge of whatever price range happened to be drawn. Every premium comes from the chain, so a link can never carry a made-up price.

Every figure carries the number of past occurrences it rests on. A base rate specific to one named event is shown only where the archive holds enough of them; below that the class's own rate is shown instead, and where both are thin nothing is shown at all — a first-of-its-kind event does not borrow a base rate.

A base rate is a count of what happened on past sessions of that kind. It is not a statement about the session that is coming, and the page states it in the past tense for that reason.

The band decides what is drawn, not what is stored, so it is a strike count rather than a percentage: a reader needs a stable number of rows that does not reflow as the index moves. Where the heaviest strike sits beyond the band it is pinned at the edge with its distance flagged rather than dropped.

The totals and the put-call ratio at the top are over the whole chain, not the displayed band. A ratio computed from twenty-one rows would be a different number from the one every other screen shows, and a reader comparing them would think one of them was broken.

Read it together with the change figure, not instead of it. A session can be almost unchanged on the day and still have travelled a full range and closed two points off its low — both statements are true, and one number alone shows only the first of them. That is the reason this panel exists.

It does not name a candlestick pattern. Naming one would carry an implied next move, and the family of claims that prior turning zones mark future reversals was tested on this archive and came back below its own random control. The panel describes the bar and stops there.

Every figure on the desk carries how it was measured and how many past cases it rests on. Where a claim was tested and failed, the failure is printed beside the panel rather than left out — eleven families of directional edge, the sector rotation's persistence, and a variance-premium structure were each tested against a random control and closed.

It holds no research-analyst or investment-adviser registration and states no view on what anyone ought to do with a position. It places no orders and holds no path to a broker that could. Nothing on it is a personal recommendation to buy or sell.

As a share. The book has typically run about ten short for every one long, so a ratio sits at 0.10 to 0.12 on a scale whose far end is 1.00 and every session draws the same stub. Stated as a percentage against the record of the year behind it — how many of the last 246 sessions were more short than this one, and what the usual share has been — the same fact becomes readable.

It is not a position at a price. The exchange publishes this once a session and in aggregate across indices, so nothing here says a book was added at a particular level.

It is a count of past sessions, not a percentile dressed as a word. The desk fetches a full year of the book for this sentence specifically, because a share on its own says nothing about whether it is ordinary.

It carries no claim about what an unusual reading has been followed by. The sentence names where the book is against its own record and stops.

A formation reads "possible double top" until price has closed through the neckline, and drops the qualifier afterwards. The neckline for a double top is the low between the two peaks, which is why the word sits at the peaks rather than on the neckline — the first version of this panel labelled it there and it read as a support level.

No measured move is drawn from it. Whether the height of a formation says how far price travels afterwards was tested on both indices at every horizon and the correlation came back at about zero, so the desk records the formation and does not project from it.

The price change is always within one contract. On the session after an expiry the front contract becomes the next month, which trades at a different price for reasons that have nothing to do with the day, so the desk looks up the previous close of the same expiry rather than of whatever was front yesterday.

A contract in its first session has no previous close of its own, so its change is blank and the row says "first session of this contract" rather than printing a number made by subtracting two different instruments.

The direction is what carries. Over the sessions measured, NIFTY opened the same way as GIFT about two mornings in three, and rather more often than that when GIFT had moved at least a third of a per cent. The size does not carry: applying GIFT's whole overnight move to the previous close missed the actual open by more than assuming no gap at all.

There is no indicated opening level on this card, deliberately. One was published and then withdrawn: measured over 782 sessions, the published pointer missed by 121 points at the mean where a flat open missed by 79. What replaced it is a band of what past mornings did, with its own count.

The vertical order is the whole point of the panel: it is the one arrangement of the morning's figures that prose cannot carry, because a sentence has no up and down. The rows are evenly spaced rather than drawn to scale — the distances are written beside them, and the picture drawn to scale is the chart immediately below.

It is not a chart. The three strike rows belong to one option contract and the panel names which expiry that is; the level rows come from the daily bars and belong to no contract at all.

Every mark on it is traceable. A level's label carries how it was arrived at — how many times price turned there, or whether the weekly timeframe carries the same price — and the candle it came from is ringed. A legend under the chart names each mark and where it comes from, in words rather than symbols.

It is not a pattern read. The triangles record that a break or a formation was detected and confirmed, and the measured record of what followed each kind sits beside them; the drawing makes no claim about what a mark means for the next session.

The percentile is what makes it comparable. Volatility at fourteen meant something different in 2013 from what it means now, so the desk ranks each reading against its own recent history rather than against a fixed number.

A high percentile says option premium is dear relative to this contract's own year. Whether that is worth acting on was tested directly — a gauge comparing what the option market was asking with what the desk had measured, on both indices, net of costs — and it failed on both sides of both indices, including on expiry day where the test carries no time-scaling assumption at all. The record of that is published.

Nearest first. Where a side has no confirmed pivot at all — which happens exactly when the index is cutting through its own lows or highs, since the extremes that matter have not yet confirmed as pivots — the panel says which substitute it is showing: the extreme of the last ten sessions, marked still forming; a level from the last weekly swing, marked as such; or the nearest confirmed level however far, with its distance stated. A side is never silently empty.

A level is a location, not a signal. Roughly half of all touches see a full day's range back the other way within ten sessions, and roughly three quarters are closed through inside the same ten. Both of those are true of the same touch, and the panel prints the counts rather than choosing between them.

The rupee is quoted as rupees per dollar, so the figure going up means the rupee weakened. That is the one series on the desk whose number runs opposite to the word, which is why the desk states the direction in words beside it rather than leaving the sign to be interpreted.

These are not Indian session figures. Each is that instrument's own most recent settled reading at 08:00 IST, and the tile names the time so a reader is never comparing an overnight print with an Indian close without knowing it.

Read it top down. The first band answers what kind of market this is; the second is the written read; the third is where NIFTY stands with its chart; then overnight, then a row of three cards — sector strength, the foreign book and what is coming. Each band is self-contained, so stopping after the first two still leaves a coherent picture.

It is not live. Nothing on this page ticks. The index figure is the settled close, or the exchange's own reading taken at 15:35 on the session that has just ended; the option chain is last night's; the overnight numbers were read at 08:00. The block at the top of the page names all three of those clocks separately, and it does so because they genuinely differ.

It is an accounting identity over the open interest as it stands right now, and it moves as the open interest moves. The curve matters more than the single number: a shallow curve means several strikes are nearly equivalent, and a sharp one means a single strike dominates.

It is not a level price is drawn to. It is computed in open-interest units rather than rupees, because the contract's lot size changed twice inside this archive and the minimum sits at the same strike either way. Where the minimum is shared by more than one strike, no strike is named at all.

Start at Market view and follow the depth links: each band on the front page links to the tab that carries the same material in full.

There is no sign-in and no paid tier. Everything on the desk is open, so no tab is locked and nothing is held back behind a plan.

The honest form of the question is "on days when NIFTY was in this band, the foreign futures book grew by this much and shrank by that much" — not "the book was added at this price". The exchange publishes one reading a session, in aggregate across indices, so the second sentence cannot be said by anyone.

Neither profile is offered without the other, and neither is a level. The loudest band in a profile is usually one with nine sessions in it, which is why the counts are on the panel and the summary refuses to name a thin band.

The percentile is the half that makes the number readable. A ratio of 0.83 on its own places nowhere; the same figure marked against its own year says whether it is ordinary. The percentile is measured strictly against the sessions before the one being ranked, never including it.

It is a description of where open interest sits, not a state of mind. The desk attaches no reading to a high or a low value, because no directional claim in this project has survived its own gate.

The word describes the daily bars. The trading range beneath it is the extreme of the sessions before the latest one, so a close outside that range is not an error — it is the most useful thing the range can say, and the card marks it in the caution colour when it happens.

It is not a direction for what comes next. Eleven families of directional edge were tested on this archive against a shuffle null with cross-index replication, and every one of them came back negative; the record of that is published. The word names the shape the bars have already made.

Position on the strip means a strike, so read it against the close mark. The heaviest call strike and the heaviest put strike are the two the desk names as walls, and a wall is named only when the maximum is a clear one: on a chain where several strikes carry the same open interest, no wall is drawn at all.

A wall is where open interest is concentrated. Whether it acts as a barrier was tested on this archive and came back negative, with one marginal cell that did not replicate — the record of that is published. The strip shows where the writers are, not what price does when it gets there.

The three timeframes are shown together rather than behind tabs, because the question a reader has is whether the day, the week and the month agree, and a control that must be clicked three times cannot answer it. Disagreement is a reading rather than a fault — a rising day inside a falling month is the ordinary shape of a pullback, and the page prints all three words rather than choosing one.

The sixty-minute and fifteen-minute timeframes appear in the chooser and are refused with a note saying they arrive with the intraday layer. Nothing on this page is intraday.

Light is the front door, because the house's own site is light. The choice is remembered in a cookie and read on the server, so the page arrives already in the face chosen rather than flashing the other one first.

It is not a per-device setting synced anywhere. There is no account on this site, so the preference lives in the browser it was set in.

The reading is a percentile rather than a number, because a VIX of fourteen meant something different a decade ago. The ranking is strictly backward looking: each session is ranked against the window before it and never including itself, or the band a day fell into would depend on what came after and the whole table would be a look-ahead in a base rate's clothing.

No claim is attached to any band. A correlation between the two series on the day is well known and changes nothing, which is why the panel measures the distance travelled afterwards instead.

The line and the bands answer different questions and can honestly disagree. A week can drift a per cent while every individual session sits inside its own range, so "fell through the week" over five sideways bands is a coherent reading rather than a contradiction. The caption states the line first because that is the quantity the picture draws.

A hollow band is not a flat session. It means the desk has no structure row for that session yet, and the caption counts how many of the five are in that state. The sessions themselves are taken from the index, so a missing read leaves a hole rather than sliding an older session into its place.

The counts under the ladder are the point of the card. On this archive, a ratio level's first touch was closed through within ten sessions about three times in four, and lines drawn at ratios that are not Fibonacci ratios at all behaved the same way — so the ratio is not what makes the level. What did carry: once a close had gone through the halfway level by half a day's range, the next level down was reached inside ten sessions in a large majority of cases, and the same held for non-Fibonacci pairs at the same spacing.

It is not a claim that the numbers work. The counts beside them say what has happened at these prices, with the number of cases each rests on, and one of those counts is the measured statement that the ratio itself carries no information.

The panels share one scale, so a small wobble and a large slide do not draw the same shape. That is the classic failure of a grid of small charts and the reason the scale is shared rather than fitted per panel.

A market that was shut on a session has no point on that session rather than a flat segment. Carrying the last value forward would draw a claim about a day that market did not trade. The first version of this panel put twelve series on one set of axes and was unreadable; it says only that markets move.

Read it as a description of the desk's own figures rather than as a source of new ones. Every number in it appears somewhere else on the page, by construction — that is what the fourth check enforces.

It is not a view on what comes next, and it is not the only paragraph the page can carry: when nothing passes the checks, when there is no key, or when the board is older than the paragraph, the page composes a plainer read from the same figures instead. That fallback is the cost of the strictness, and it is never a gap.